An Algorithm Unrolling Approach to Deep Image Deblurring

Abstract

While neural networks have achieved vastly enhanced performance over traditional iterative methods in many cases, they are generally empirically designed and the underlying structures are difficult to interpret. The algorithm unrolling approach has helped connect iterative algorithms to neural network architectures. However, such connections have not been made yet for blind image deblurring. In this paper, we propose a neural network architecture that advances this idea. We first present an iterative algorithm that may be considered a generalization of the traditional total-variation regularization method on the gradient domain, and subsequently unroll the half-quadratic splitting algorithm to construct a neural network. Our proposed deep network achieves significant practical performance gains while enjoying interpretability at the same time. Experimental results show that our approach outperforms many state-of-the-art methods.

ViterbiNet: A Deep Learning Based Viterbi Algorithm for Symbol Detection

“…Our design is inspired by deep unfolding, which is a common method for obtaining ML architectures from model-based iterative algorithms [5], [20]. Unfolding was shown to yield efficient and reliable DNNs for applications such as sparse recovery [5], recovery from one-bit measurements [6], matrix factorization [21], image deblurring [22], and robust principal component analysis [23]. However, there is a fundamental difference between our approach and conventional unfolding: The main rationale of unfolding is to convert each iteration of the algorithm into a layer, namely, to design a DNN in light of a modelbased algorithm, or alternatively, to integrate the algorithm into the DNN.…”

Deep Learning in Ultrasound Imaging

“…To overcome these limitations, in [81], [82], [83], the task of clutter removal was formulated as a convex optimization problem by leveraging a low-rank-and-sparse decomposition. The authors of [81] then proposed an efficient deep learning solution to this convex optimization problem through an algorithmunfolding strategy [84]. To enable explicit embedding of signal structure in the resulting network architecture, the following model for the signal after beamforming was proposed.…”

Gaussian Kernel Mixture Network for Single Image Defocus Deblurring

“…Most existing optimization unrolling based image deblurring methods (e.g. [21,22,20,23,24,25,26,27]) consider uniform blurring, where the matrix B can be represented by a convolution. The iterative schemes they adopt such as HQS, usually involve an inversion process for B, which can be efficiently computed using FFT when B is a convolution operator.…”

“…[20,23,26]. There are also some works [21,22,24,27] on unrolling-based blind image deblurring where the PSF is unknown. However, these methods are restricted to the case of uniform blurring.…”